The AI Prompting Playbook: Writing (Part 1)

Everyone uses AI to write now. Blog posts, emails, marketing copy, social captions. The tools are everywhere.

But here's what happens in practice: You open Claude or ChatGPT, type a prompt, hit enter, and get back something that sounds... fine. Generic. Flat. Clearly AI-generated.

So you try again. Tweak the prompt. Still not right. Eventually you either give up and write it yourself, or you spend 30 minutes editing AI output that was supposed to save you time.

The problem is the prompt.

Most people prompt like they're ordering fast food: "Write a blog about productivity." That's the equivalent of walking into a restaurant and saying "give me food." You'll get something. It probably won't be what you wanted.

This guide teaches you how to prompt for writing that sounds good on the first or second try, not the tenth.


Why Most AI Writing Sounds Like AI

Before fixing your prompts, understand why the default output is so mediocre.

AI writes to the middle. It's trained on millions of documents, so it defaults to the most common patterns. That means corporate buzzwords, safe sentence structures, and zero personality. The output is average. And average is forgettable.

AI doesn't know what you know. When you write "write a blog about remote work," the AI has no idea who you are, who's reading, what angle you want, or what tone fits your brand. So it guesses. And guesses are generic.

AI takes you literally. If you ask for "a blog post," you get a blog post. If you want something specific (a punchy intro, a controversial take, a first-person story), you need to ask for it. AI won't read your mind.

The fix is better input.


The 5 Elements of a Good Writing Prompt

Every strong writing prompt includes some combination of these five elements:

ElementWhat It DoesExample
RoleTells AI who to write as"You're a SaaS founder writing for other founders"
ContextBackground the AI needs"This is for a LinkedIn audience, mostly CTOs and tech leads"
TaskWhat you want, specifically"Write a 150-word post about the hidden cost of technical debt"
FormatStructure, length, style"Use short paragraphs. No bullet points. Conversational tone."
ExamplesShow what good looks like"Here's a post I liked: [paste example]. Match this style."

You don't need all five every time. But the more specific you are, the less editing you'll do later.

Let's see how this plays out across different writing types.


Blog Posts and Long-Form Content

The Common Mistake

Write a blog post about AI tools for developers.

This prompt will give you 500 words of surface-level content that says nothing new. It'll have a generic intro ("In today's fast-paced world..."), obvious points, and a forgettable conclusion.

The Better Approach

Don't ask AI to write the full blog in one shot. Break it into steps:

Step 1: Generate an outline first

I'm writing a blog post for software developers about AI coding assistants.

Angle: Most developers use AI wrong. They accept the first output instead of iterating.

Target reader: Mid-level developers who've tried Copilot or Claude but feel underwhelmed.

Give me a blog outline with 4-5 sections. Each section should have a clear point, not just a topic.

Step 2: Expand section by section

Once you approve the outline, expand each section individually:

Expand section 2: "Why accepting the first output fails"

Keep it practical. Include one specific example of a prompt that developers commonly use and why it produces mediocre code. Then show how to improve it.

Tone: Direct, no fluff. Like you're explaining to a colleague, not writing a textbook.

Length: 200-250 words.

Why this works: You maintain control. Each section gets full attention instead of being rushed to fit a word count. And you can course-correct before the AI goes in the wrong direction.

Template

I'm writing a blog post about [TOPIC].

Angle: [YOUR SPECIFIC TAKE OR ARGUMENT]

Target reader: [WHO THEY ARE AND WHAT THEY ALREADY KNOW]

First, give me an outline with [NUMBER] sections. Each section should make a specific point, not just introduce a topic.

Emails: Professional and Outreach

The Common Mistake

Write a professional email to follow up on a meeting.

You'll get something that sounds like it was written by a corporate robot. "I hope this email finds you well. Per our previous conversation..."

The Better Approach

Give the AI context about the relationship, the goal, and the tone:

Write a follow-up email after a sales call.

Context:
- I spoke with Sarah, Head of Engineering at a fintech startup (50 employees)
- We discussed their struggle with scaling their backend team
- She seemed interested but mentioned budget concerns
- The call was friendly and casual, first-name basis

Goal: Get a second call scheduled to discuss our dedicated team model

Tone: Warm but professional. Not salesy. Acknowledge the budget concern without being pushy.

Length: Under 150 words. No fluff.

Why this works: The AI now knows who Sarah is, what she cares about, and how you want to sound. It can write an email that feels personal instead of templated.

Template

Write a [TYPE] email to [RECIPIENT + THEIR ROLE].

Context:
- [RELATIONSHIP: How you know them, previous interactions]
- [SITUATION: What happened, what they care about]
- [CONSTRAINTS: Any objections, concerns, or limitations]

Goal: [WHAT YOU WANT THEM TO DO]

Tone: [HOW IT SHOULD SOUND]

Length: [WORD COUNT OR "KEEP IT SHORT"]

Marketing Copy: Ads and Landing Pages

The Common Mistake

Write copy for a landing page for my project management tool.

You'll get feature-focused copy that sounds like every other SaaS website. "Streamline your workflow. Boost productivity. Collaborate seamlessly."

The Better Approach

Marketing copy needs to understand the audience's pain, not just the product's features:

Write landing page headline and subheadline for a project management tool.

Target audience: Startup founders (5-20 employees) who've tried Asana or Monday but found them too complex for their small team.

Pain point: They waste time setting up elaborate systems that nobody uses. They want something simple that works out of the box.

Our differentiator: Dead simple. No learning curve. Set up in 5 minutes, not 5 hours.

Tone: Confident, direct, slightly playful. Not corporate.

Give me 3 options for headline + subheadline combos.

Why this works: The AI writes to the pain point ("too complex," "nobody uses it") instead of listing features. It knows the comparison point (Asana, Monday) and the audience's mindset.

Template

Write [ASSET TYPE: headline, ad copy, landing page section] for [PRODUCT].

Target audience: [WHO THEY ARE + WHAT THEY'VE TRIED BEFORE]

Pain point: [THE PROBLEM THEY'RE TRYING TO SOLVE]

Our differentiator: [WHY WE'RE DIFFERENT/BETTER]

Tone: [VOICE AND STYLE]

Give me [NUMBER] options.

Social Media Captions

The Common Mistake

Write a LinkedIn post about the importance of company culture.

Every platform has different norms. LinkedIn isn't Twitter. Instagram isn't LinkedIn. AI doesn't automatically know which platform you're writing for, so it defaults to generic.

The Better Approach

Specify the platform, format, and voice:

Write a LinkedIn post about why I stopped doing annual performance reviews at my company.

Context: I'm a founder of a 30-person software agency. We switched to monthly 1:1s instead and saw better results.

Format:
- Hook in the first line (make it stop the scroll)
- Short paragraphs (1-2 sentences each)
- End with a question to drive comments
- No hashtags
- Under 200 words

Tone: Personal and opinionated. First person. Like I'm sharing a lesson learned, not lecturing.

Why this works: You've defined the platform norms (short paragraphs, hook first, question at end), the voice (personal, opinionated), and the constraints (under 200 words, no hashtags).

Template

Write a [PLATFORM] post about [TOPIC].

Context: [YOUR BACKGROUND AND WHY THIS MATTERS TO YOU]

Format:
- [STRUCTURE: Hook, paragraphs, CTA]
- [LENGTH]
- [WHAT TO INCLUDE/EXCLUDE]

Tone: [VOICE AND STYLE]

The Iteration Trick

Even with good prompts, the first output isn't always final. But don't start over. Iterate.

Bad: Deleting everything and writing a new prompt from scratch.

Good: Refining what you have.

Useful follow-up prompts:

  • "Make the intro more direct. Cut the first two sentences."
  • "This sounds too formal. Rewrite in a more conversational tone."
  • "Keep the structure, but make each point sharper. Less explanation, more punch."
  • "The ending is weak. Give me a stronger closing line that ties back to the opening."

Think of it like editing a draft with a co-writer. You don't throw away the whole draft because one section is off.


Prompt Templates to Steal

Blog Post

Template:

I'm writing a blog post about [TOPIC].
Angle: [YOUR SPECIFIC TAKE]
Target reader: [WHO + WHAT THEY KNOW]
Give me an outline with [NUMBER] sections. Each section should make a specific point.

Example:

I'm writing a blog post about onboarding remote developers.
Angle: Most companies over-document and under-communicate. The fix is better async rituals, not more docs.
Target reader: Engineering managers at startups (20-100 employees) who've struggled with remote onboarding before.
Give me an outline with 5 sections. Each section should make a specific point.

Email

Template:

Write a [TYPE] email to [RECIPIENT + ROLE].
Context: [RELATIONSHIP, SITUATION, CONSTRAINTS]
Goal: [DESIRED ACTION]
Tone: [VOICE]
Length: [WORD COUNT]

Example:

Write a follow-up email to James, CTO at a healthcare startup.
Context: We had a discovery call last week. He's interested in outsourcing their mobile app development but worried about communication with offshore teams. He mentioned they tried outsourcing once before and it failed due to timezone issues.
Goal: Book a second call to discuss how we handle async communication and overlap hours.
Tone: Friendly, reassuring, not pushy. Acknowledge his past experience.
Length: Under 120 words.

Marketing Copy

Template:

Write [ASSET TYPE] for [PRODUCT].
Target audience: [WHO + WHAT THEY'VE TRIED]
Pain point: [THEIR PROBLEM]
Our differentiator: [WHY WE'RE DIFFERENT]
Tone: [VOICE]
Give me [NUMBER] options.

Example:

Write a hero section headline and subheadline for a invoice automation tool.
Target audience: Freelancers and small agency owners who currently use Excel or Google Sheets to track invoices.
Pain point: They forget to follow up on unpaid invoices and lose money. Manual tracking is tedious and error-prone.
Our differentiator: Automatic payment reminders. Connect once, never chase payments manually again.
Tone: Friendly, straightforward, no corporate jargon.
Give me 3 options.

Social Media

Template:

Write a [PLATFORM] post about [TOPIC].
Context: [YOUR BACKGROUND]
Format: [STRUCTURE, LENGTH, INCLUSIONS/EXCLUSIONS]
Tone: [VOICE]

Example:

Write a LinkedIn post about why I hire for curiosity over experience.
Context: I'm a founder of a software development agency. I've hired 40+ engineers over 6 years. Some of my best hires had less experience but asked better questions in interviews.
Format: Hook in first line, short paragraphs (1-2 sentences), end with a question, no hashtags, under 180 words.
Tone: Personal, reflective, like sharing a hiring lesson. Not giving advice.

What's Next

Prompting is a skill. The more specific your input, the less you'll fight with the output.

But writing is only one use case. In Part 2, we'll cover research prompting: how to get AI to find, verify, and synthesize information without hallucinating or giving you shallow summaries.

Same framework. Different application.


Feb 03, 2026

Similar articles

AI Code Review Is Broken: The 4 Layer Model That Fixed It for Our Team

63% of developers spend more time debugging AI-generated code than writing it themselves. We built a 4-layer review model that gives AI a specific job and keeps humans focused on what AI can't judge.

Top AI Agents for Sales & Customer Support in 2026

Stop handling every support ticket and sales email manually. 6 AI agents compared by pricing, features, and which one fits your team best in 2026.

The AI Prompting Playbook: Researching (Part 2)

AI research fails when you prompt for answers instead of process. Learn the 4-element framework and templates for research that's accurate and verifiable.